Green IO
#34 - Carbon aware computing: a false good idea? with Hannah Smith and Ismaël Velasco
March 12, 2024
🎙️In our latest episode, Hannah Smith, the COO of the Green Web Foundation, and Ismael Velasco, the founder of the Adora Foundation, highlight the potential pitfalls of Carbon Aware Computing a strategy that aims to reduce carbon emissions by aligning software operations with the carbon intensity of the energy grid. While this approach has gained traction among the IT industry, including giants such as Apple, Microsoft, and Google, Hannah and Ismael discuss with our host Gaël Duez its complexities and limitations. 💻As they explored the nuances of Carbon Aware Computing, it gave us insights that while it holds promise as a tool for reducing emissions, it also poses challenges that must be carefully considered. The oversimplification of energy grid dynamics and the myriad factors that influence carbon intensity underscore the need for a more comprehensive understanding of sustainable computing practices.
🎙️In our latest episode, Hannah Smith, the COO of the Green Web Foundation, and Ismael Velasco, the founder of the Adora Foundation, highlight the potential pitfalls of Carbon Aware Computing a strategy that aims to reduce carbon emissions by aligning software operations with the carbon intensity of the energy grid. While this approach has gained traction among the IT industry, including giants such as Apple, Microsoft, and Google, Hannah and Ismael discuss with our host Gaël Duez its complexities and limitations.

💻As they explored the nuances of Carbon Aware Computing, it gave us insights that while it holds promise as a tool for reducing emissions, it also poses challenges that must be carefully considered. The oversimplification of energy grid dynamics and the myriad factors that influence carbon intensity underscore the need for a more comprehensive understanding of sustainable computing practices.

❤️ Subscribe, follow, like, ... stay connected the way you want to never miss our episode, twice a month, on Tuesday!

📧 Once a month, you get carefully curated news on digital sustainability packed with exclusive Green IO contents, subscribe to the Green IO newsletter here.

📣 Green IO next Conference is in Singapore on April 18th (use the voucher GREENIOVIP to get a free ticket) 

Learn more about our guest and connect: 


 
📧 You can also send us an email at [email protected] to share your feedback and suggest future guests or topics.   
 

 Hannah and Ismaël's sources and other references mentioned in this episode: 

 

Transcript

 
 
[00:00:00] Gaël Duez: Hello, everyone. Welcome to Green IO with Gaël Duez. That's me. Green IO is the podcast for responsible technologists building a greener digital world, one byte at a time. Twice a month, on a Tuesday, our guests from across the globe share insights, tools and alternative approaches enabling people within the tech sector and beyond to boost digital sustainability. And because accessible and transparent information is in the DNA of Green IO, all the references mentioned in this episode, as well as the transcript, will be in the show notes, both on your podcast platform and on our website, greenio.tech. Let me share with you a personal story. Last July, I was on a rooftop in Paris, sharing a drink with a friend who happens to be a former colleague of mine.

One of the most talented CTO I had the chance to work with. Loïc was asking about what I was doing in the green IT field and the quick wins I was advising to CTO willing to ramp up sustainability. I mentioned carbon computing as a no brainer and I was ready to move forward when he posed me and asked me what I was referring to, POP, do you hear that sound? Yes, another information bubble just exploded. I was so sure that almost every CTO knew the concept, I explained it, he was enthusiastic about it, and since then, it has been my magical key to onboard CTO and Cloud Ops, folks. It's easy to grab, pragmatic, and actionable. They love it. And they're not the only ones.

Almost all cloud providers launch some kind of carbon-aware program. And it's a pillar in the Green Software Foundation course, for instance, to get the Green Software for Practitioners certificate. But a few weeks ago, I stumbled upon a remarkable series of articles written by Hannah Smith, based on initial research and insights by Ismael Velasco, about some serious pitfalls with Carbon Aware Computing. It might be another case of micro-optimizations not aligning with macro optimizations, like when someone thinks that it's a great idea to film a live show with a smartphone to get souvenirs, but actually, everyone does it and no one can see the stage anymore. I have been lucky to have already had both Ismael and Hannah on the show, so it was super easy to have them back for an in-depth discussion about Carbon Aware Computing.

[00:02:28] Gaël Duez: A quick reminder for the people who didn't have the chance to listen to Green IO episode 26 or episode 5 about Hannah and Ismael. Hannah Smith, based in the UK, is currently the Chief Operating Officer of the Green Web Foundation. She's also a pillar in the web WordPress sustainability community as well as a long-time volunteer for the climateaction.tech community and overall an amazing community builder. Hello Green Tech South West as well as a season green software practitioner. Ismael Velasco is also a member of the amazing community. He's based in Mexico and he's a true veteran in the software industry at large and green software in particular. You should check his API grateful degradation concept. His knowledge and commitment go far beyond Green IT with his involvement in the Adora Foundation, which fosters social innovation across the globe, and recently decided to focus more on the sustainability side of technology. So, welcome Hannah, and welcome Ismael, I'm so glad to have you back on the show.

[00:03:35] Hannah Smith: Thanks, Gaël. I knew it was a while ago that we recorded our episode. I didn't realize it was episode five. 

[00:03:44] Gaël Duez: Wow. Yeah, that was a while ago. I don't know why I've waited so long to bring some of my former guests back to the show, but I think I will do it more and more again. The pleasure of having you both here is great. So thanks a lot. And before we deep dive into both your approaches, maybe it's time to set the stage, I would say. And I would, I'd love to ask you, Hannah, the first question, which is not that easy maybe, what is Carbon Aware Computing? 

[00:04:12] Hannah Smith: Well, oh, I don't know. I think we can make it easy for people to understand. At its basic level, there are two ways that you can approach Carbon Aware Computing. We do talk about it in the article that we wrote together. There is the idea of time-shifting software, and there's the idea of location-shifting software. And you do both of those things based on the carbon intensity of the grid at any particular time. So what that implies is that maybe you've got an API that you can access that gives you data about what the carbon intensity of a grid is at any given time. 

For carbon intensity, if anyone's not sure what that means, that simply means how clean or dirty energy is. So you have a high carbon intensity if energy is dirty, e.g. is being produced a lot with fossil fuels, and you have a low carbon intensity if it's renewable produced energy, so it's not emitting as much carbon. So the idea of carbon awareness is looking at the carbon intensity of energy and then shifting your software around, either to run at different times of the day or to run in different physical geographical locations. I think in a nutshell, that's it. I know Ismael, you were involved in writing the, well, I think the very first post about Carbon Aware on, Hackernoon. So I wonder what else would you add to that description? 

[00:05:38] Ismaël Velasco: Yes, I think that is, there's not much to add really, because it is quite a simple concept. So simple and so kind of commonsensical. It was surprising for us to question it because it just made so much sense the grid is mixed, so when is it greener? It's greener. I run my computer there. Therefore, I'm greener that's a simple, principle, right? If the grid has more renewables, I time my computing to run when the green, grid is greener. And that should mean that my computer is reducing emissions because it's running on that green. And that is the common sense idea that's growing. And maybe just to give some examples, so people can. Envisage it more concretely. Apple released an update for their American US iPhone, where you keep it plugged in, and it only charges when the grid is particularly green in your area. So if the grid is mostly powered by carbon and fossil fuels. Your iPhone will not run electricity, but when the grid is particularly green, then it will charge your phone. And that way it's been greener. Microsoft has done the same for your Windows updates. So your computer will wait until the electricity is greener in your location. And if it knows that it's currently greener, then it will say, update my computer now, instead of doing it, when the electricity is dirty. So those are types of examples of how you might do that. And the theory of change is that by doing that, you are avoiding emissions, which is the theory of change that Hannah and I questioned.

[00:07:31] Gaël Duez: And the examples you provided, Ismael, are both time shifting, I reckon that quite a lot of time we hear also the expression chasing the sun or chasing the wind, so do you have also examples of location-based Carbon Aware Computing? 

[00:07:50] Ismaël Velasco: The people who have done this at the biggest scale are Google. So Google first implemented it internally, and now they offer it to all the users of their cloud. So they started computing, where all their data centers across the world were. And then they started finding out, in which location on the planet, the electricity was greenest. And then if they were to do say a backup and could run that backup from a server in any of 50 countries. They would choose a country that had the greenest energy at that particular time. So instead of just saying, I'm going to wait until it's the right time, like with the Apple phone, they would say, I want to do it right now, but I'm going to run it somewhere in the world where it is greenest. And now if you go, and you have your applications hosted in the Google Cloud you can see Which server is currently greenest and there's an API you can even see and say this one is greenest so you can say, okay, whenever you run a job in my application look in Google and find the location. Chase the sun, find the location where? There's a lower electricity and do it there. So we'll put all our jobs there and again, theory of changes. That way we're reducing emissions. 

[00:09:19] Gaël Duez: Do you know if the prices across the region vary accordingly I had a very interesting discussion with a cloud ops people who wanted to implement it with another big hyperscaler. Let's not name it. And they discovered that the bill would go up by like 30%. So when you see Google went full speed on this Google Cloud Platform, went to full speed on this solution, do you know if it costs, people to move from one place to another, or if you chase the sun or the wind, actually you don't see an increase or at least a significant increase in your bill?

[00:09:57] Hannah Smith: That's a great question. I didn't, I couldn't answer that confidently and say, I mean, generally, you know, the renewable energy, on the whole, is cheaper, but I don't know if that necessarily equates in the way that you're saying it, Gaël. That's a great question. An interesting one. 

[00:10:13] Ismaël Velasco: My understanding is that generally, it is cheaper because very often It's not always this is one of the assumptions that people kind of take for granted But it is very common that the times when the energy is greenest is also when demand is lower. So It can often be cheaper countries like the UK are beginning to roll out carbon pricing as well around electricity and this is something that's growing. So it varies from place to place there have been places where everybody piling on to green energy, so this happened in the US, for example, they started in Texas, among other places, but especially in Texas, some people install US miners in the electricity plants when the electricity was dirty and demand was high, they were paid not to run any compute. And when the demand was low, they were paid to run it. And it was when it was greenest. The result was the electricity prices for the city rose significantly because US used so much. So it's not an absolute rule, but in general and increasingly targeting the greener energy will often be slightly cheaper.

[00:11:34] Gaël Duez: I'll try to research that and put, the answer in the show notes before the episode is released. And I think we, we are good. Because it's a very straightforward concept. It is getting adopted very widely, with several examples from very big tech companies. So, problem solved, and we've got, a word record of the shortest, Green IO episode ever. So is the problem solved and that is the best possible tool to use, to reduce carbon emissions from the grid, or actually do we have a problem? What is actually the issue with Carbon Aware Computing? 

[00:12:15] Hannah Smith: Yeah, well, it's an interesting question because as Ismael said, right, you explain it in the way we've just explained it, and you're like, oh my god, this is the best solution ever. Of course, we should be doing this. This is like, amazing and so logical. So we were discussing this in climateaction.tech, which we're both, as you mentioned, Gaël, in the beginning, we're both members of, and we both hang around in the Slack group. And Ismael posted this thing saying, Hey, look, I've been kind of looking at this carbon-aware stuff. And when I look really deeply at this, and when I actually think about how the grid works in practice, Ismael was like, I'm concluding that this is maybe not the solution it seems. And Ismael in his fantastic way was really in-depth, loads of examples, kind of really kind of got into the problem. And, I happened to stumble upon it at the time that Ismael posted it. And I was like, Oh my word, I think he's got a bit of a point here. This is actually a little problematic. And it all comes down to this idea that carbon aware. Is looking at a very simplistic metric, which is the carbon intensity of energy, at a given location or at a given time, but it's doing that in isolation from how the grid actually works in practice, which is phenomenally complicated and has a lot of nuances around it. And what Ismael did was kind of present the issues and present some concerns here. So I was quite interested in that and sort of said to Ismael, Hey, let's work on this together. Let's tidy this up. Let's format it. And, let's really explain to people in, in as much depth as we need to go into why we think there's a problem here, but in a nutshell, grid aware.

Sorry, carbon aware is not thinking about the realities of managing the grid, of how supply and demand actually work in practice. And what happens when you suddenly just shift a load of stuff from one place to another. So yeah, I think there is a problem. But I think it's really important to say that it's not a problem where it means that the whole of this concept is nonsense and shouldn't be done. I think that that was something Ismael and I wanted to clarify upfront. There's a real benefit here. There's real potential for awesomeness but not in the way it's being done at the moment. So I think the way we turned to Ismael was we talked about these warning labels. We said, Hey, look, there's these warning labels that aren't being applied to these approaches at the moment.

There are these nuances that are just being conveniently overlooked and this approach is just a little bit too simplistic at this point in time. But you know, the awesome thing is, it can be matured, it can iterate, it can improve, and that's really what we're, well, what we set out to try and do with this work. And this proposal we came up with on Grid Aware.

[00:15:16] Gaël Duez: Ismael, could you enlighten us a bit on that, how the grid works? And then explain to us what are the, the pitfalls or the aspects of the Carbon Aware Computing concept, which are conveniently, overlooked, and Hannah just said. 

[00:15:34] Ismaël Velasco: Yes. And first I just wanna share one of the kind of subjective elements of this. I began super enthusiastic about this. I sort of worked with the Green Software Foundation to organize the first carbon aware hackathon and, sort of looked at their APIs. I love that. I evangelize, I move people, et cetera. It was great. Then as Hannah said, I've got some confusion here. And the biggest question is actually, again, really simple, which is what made me so suspicious of my own questions. If I run my software, if I consume energy when the grid is very green, am I taking away emissions from the planet? Now, that's weirdly enough the question that no one seems to actually have asked themselves at all. And the key concept is very logical, is that if we were running on 100 % or even 99 % green energy, that might work. But at the moment, the planet has a supply of 40 % of renewable energy, 43, and a demand of 100%. And in every country, maybe not Iceland. It's fractional, right? It's 40, it's 50, it's 20, it's 70. But the point is that every day and every year and every five years, there will be a 100 % amount of electricity used. And the emissions of that 100 % will not come from the green energy, they will come from the extra, right? That's logical. Now, the analogy that helped me clarify this is if you think of the electricity we use each day, each year, pick your timeline as a train with 10 carriages, each carriage takes 100 passengers and each passenger is an emission.

Four of those carriages are green, and six of those carriages are dirty. What is Carbon Aware Computing? Imagine those four carriages appearing randomly on the train. On the first day, they're all at the front, on the second day one is in the middle, but there are always four carriages out of ten. Now, most days, I come in at eight o'clock, I go to the train, and I go into the first carriage in front of me. It might be green, it might be dirty. But now I'm going to do carbon awareness. I'm going to make sure that every day I take the green carriage, no matter what, I will move to the place where the green carriage is. I will tie my arrival and I will always go on the green carriage that provides zero emissions. I'm going to come out of that green carriage feeling so clean. I've been clean all week. Every day I've been in a green carriage. My body is absolutely clean. Perfect. But what happens? That carriage is always full. So the day that I came into that green carriage, the person who always comes into that green carriage. Now can't get in because I'm already there and there's only room for 10 people. So she has to go into one of the dirty carriages In other words, you always have whenever I get on the plane or the train There will always be 400 people in dirty carriages. So I'm feeling very clean but the emissions. The total emissions of the train are the same regardless of when I run my computer I will always run it In that maximum 40 percent green period there are only a number of exceptions. How could you do this and actually reduce the number of passengers that come out of the dirty carriages?

One is if I don't go on the train. Right? If I reduce a person, I don't go on the train, now there are not 1, 000 passages, there are 999. I've reduced emissions. So if I can reduce the electricity that my computer uses, great. That has been great. The second one is if I use electricity that is going to be thrown away, it's called curtail energy. Imagine that the wind blows really hard in Scotland at 3 a.m. That's more energy than the grid can use. So they throw it away. If I run my job there. Then that's not going on the train, right? Those are people outside the train. I am reducing my emissions but the problem is that that is I've estimated between one and three percent of all global usage. So you'll never run at most you're increasing our green percentage to 46 % it's meaningful but time shifting still not really reducing emissions, and it's not straightforward 95 % of curtailment in Britain happens in Scotland. So we only have 5 percent of curtailed energy. In other words, it could be good if you can time it, but it's not enough to make a really big difference. 

[00:21:28] Gaël Duez: If I had to sum it up, it would be. That we have a limited amount of, low carbon electricity supply. And if you get some strategy to have access to it, no matter the time, no matter the day, like optimizing, like hell, if we don't manage to either increase the total amount of, I would say clean energy supply to follow your wording or that we don't manage to reduce the amount of demand that will by default might go to a high intensity, high carbon intensity energy. Then it's, yeah, it's a zero-sum game. 

[00:22:18] Ismaël Velasco: Brilliant. Absolutely. Yeah.

[00:22:21] Hannah Smith: I mean, I think one of the key things to think about here with this, something that really helped me is I really liked Ismael's carriage analysis. I found that really helpful. I also found it helpful to think about it like a balloon. The balloon is still the same size, and all you're doing is just squeezing air into different parts of it. You're not actually reducing the amount of air in the balloon, you're not actually making a meaningful difference to the emissions coming out. And that is down to the fact that the grid always has to balance supply and demand. And that's this element of grid management that isn't really taken into account with this approach, is those people managing the grid have to always keep supply and demand in balance. And as Ismael was saying, what they'll do is they'll use as much of the renewable energy as they can, but then they'll top it up with fossil fuels to meet that 100% of demand. And that's always happening, that's always the case. And they create that supply and demand equation through data, through looking at typically when there is a certain amount of demand, so you know that there's more demand during the day, there's less demand during the night, and they forecast that and schedule it and manage the outputs of all the different power plants in the grid to meet that demand. So you're coming along and making a sudden change doesn't actually make a difference to what those grid operators are doing, because it's unpredictable. They don't know it's going to happen, and it doesn't materially make an impact. And what Ishmael said, I'll just say this bit, because I think this is a useful bit to build on as well. What Ismael helped me realize, and what we talked about a lot within the article, was that, furthermore, shifting that demand unpredictably can actually ramp up the amount of fossil fuels going on the grid because the demand suddenly increases. All right? And look, at the moment, we're talking about fractions of percent. So it's not manifesting in a mad way. But one of the things that we were worried about is that if carbon aware is done at scale, what you actually find is grid operators are then bringing more fossil fuels online because they're the quick things to ramp up to meet the demand. So it doesn't really seem to make sense at the moment when you think about that reality of things going on.

[00:24:46] Ismaël Velasco: And that's a really important point because it means we have two dimensions to our argument, really. One is, don't assume it's helping. Right? That's the first one. And what you wrote is exactly what the White House report on crypto summarizes. Said the only two ways you reduce your emissions through computing is if you reduce your demand or you increase the green supply. Most carbon aware patterns don't do either, so they're not affecting it. So, one is, are you making a positive difference? The second point that Hannah just introduced is that there's also the question that no one seems to be asking, which is, are you making a negative difference? And what we found, Hannah and I, as we started looking at it, is that there are lots of scenarios where running carbon aware patterns at scale could not only not help, but it could actually create harm. And we've got some examples of this happening. In Iran, for example, bitcoin miners used electricity so much at the same time that the whole grid broke. The same in Venezuela. So you could imagine that if you happen to have the greenest energy at exactly this moment in a place with a creaky grid, like Texas or Azerbaijan also collapsed. And all of Google says green everybody and runs everything there. You could imagine that you might bring the grid down.

[00:26:35] Gaël Duez: But playing a bit of the devil's advocate here, I would have two other questions. The first one is, what about market incentivization? So, like, yes, at an instant time, you've got this demand and supply balance, which might lead to not generating a positive impact, even maybe a negative impact. But the adoption of more and more carbon aware computing practices will drive cloud providers, software providers, you name it, et cetera, energy providers, et cetera, to adopt greener practices more generally to supply more low carbon energy. I would say that will be my question number one. My question is number two, could it also be an incentive to avoid curtailment? And I would like, Ismael, if you could maybe explain a bit more curtailment and the example you took between Scotland, to England, for instance. Because what I understood, actually quite a fun fact, is something that we discussed, just the latest episode on open source software, about all to connect which we are building, at least in Europe, but in many other parts of the world, to connect the grids and to avoid curtailments, to avoid wasting the most precious thing that we've got today, which is low carbon electricity. Could it be an incentive also to make sure that we don't waste energy so we reduce dramatically curtailment? My two questions could be summarized as what about the market incentivization in the medium term, not even the long term? And what about the possibility of using in a better way wasted energy today, mostly curtailed energy?

[00:28:28] Ismaël Velasco: So I'll give my version of a very brief answer, but I also think because of time, that this is also a good opportunity for Hannah to introduce the idea that we're not just condemning carbon aware software, that we're saying, let's do it properly, let's do it well, that there are improvements and that this could be a really good solution. But we do need to ask those questions, so it's not all doom and gloom. So one of the positive arguments for carbon aware patterns is that one around incentivization, around culture. So, for instance, I got super excited by carbon aware software and it motivated me to dive into all kinds of other things. I eventually arrived at the conclusion that, oh, I think there's a problem here, Houston. We've got a problem like Hannah put it, but thank goodness that I got into it because it opened all kinds of other fields. So that's great. The other argument that is made in favor of it is that by investing and timing it to the green times, you are motivating the market to invest more in renewable energy. And I think that was true a while ago, but I think, weirdly enough, we won that argument a while ago. The market for renewable energy is now cheaper, so people are investing. The news came out today that there are about four or five countries in Europe that have hit their 2040 or 2030 target now, not because of incentives around carbon, our computer, it's just cheaper. But I think there is a cultural incentive around anything that we do intentionally for green issues, for transparency, for clarity, and evidence. So there are ways to do this. And you can say at the end of the day, I reduced my demand. I used new green energy, like curtailed energy. But you've demonstrated that you've asked yourself, and you've also asked yourself, is this having any perverse side effects? No. And I've said it. Google says it, and Microsoft says it. Right when they're doing this at scale, Apple says it. But when nobody says it, then the potential for greenwashing and perverse effect is huge, especially since this is one of the few areas where the big corporations and the really sincere green advocates like ourselves currently are on the same page. So it's very easy for everyone to go, yay. So we need to simply say, do it, but tell us how it's helping. One example of that, just to finish and answer your question, is curtailed energy, which is basically this idea of green energy that is happening, but it's being wasted because it's happening at the wrong time.

[00:31:24] Hannah Smith: So I think the question of curtailed energy is a really good one, and I think it's a really awesome incentive. And through the research, and there's this really deep level of thinking that we did on this, we kind of realized that using curtailed energy is one example where Grid, carbon aware kicks ass. It really is like awesome. And that's what we want to incentivize. It makes sense. And it seems to stack up. So when we were writing this post, as Ismael said, we didn't want this to be all doom and gloom. And we didn't want this to be a takedown because it's too easy to take down other people's work. And that's not really how we build sustainable solutions. We want to build each other up and, you know, work, work, work, work, work, work on each other's stuff. So what we did, we were like, okay, we need to come up with a kind of name and way of talking about. The Times When Carbon Aware Does Work. It's a working title, but we came up with the idea of talking about grid aware software instead. And so that really, the idea of that naming was to say kind of, hey look, this is more than just thinking about this simple metric of carbon intensity. This is actually thinking about the grid and all the complexities there. Because actually, that seems to then push you in a more nuanced direction with the solutions.

And so what we said was one of those examples, as I mentioned, is using curtailed energy. But not just curtailed energy in its simplistic form, but coming back to something Ismael was talking about earlier, it's using curtailed energy in stable grids in places which can actually absorb this additional demand coming in. And on top of that, I think it's also doing it in a way that is in harmony with the grid. So I think that long term, what we need to start seeing, is APIs and standards and conventions so that data centers can actually agree to use that data, and they can actually, sorry to use that energy. And there can actually be handshakes that, Hey, we've got curtailed energy. Hey, then data center over here. Now is your chance to be helpful to the grid, and to get some demand. You know, get some computer running here and make use of it. So rather than it being like a free for all pile on, I just, when Nishma was talking earlier, I had this like the image of American football where they all, or rugby, where they all just kind of pile on each other, and it all just goes disastrous, you know, and this, this horrible scrum, and it's all just gone tits up. Like, that's not what we want. We want it to be kind of done in a more organized, democratic fashion. Where the curtailed energy, like someone in the grid is saying, Hey, it would be really helpful if right now someone could come along and use that curtailed energy. And then we have these standards and processes in place that people can know that that's available and send their compute there. And that section's missing at the moment. So Gaël, you're right. This question about curtailed energy feels really central to this proposition of making this load better. 

[00:34:37] Gaël Duez: How operational today is a Grid Aware computing approach?

[00:34:42] Hannah Smith: I think there's work to be done here. So in sharing this work, it's popped up all sorts of interesting people discussing this. I mean, I can talk about the UK. That's where I'm based. It's probably the grid that I have the most hands-on knowledge about. Ismael already mentioned, that up in Scotland, we know that that's where the majority of curtailed energy is. Now, if anyone's not familiar with the geography of the UK, Scotland's right up in the North. It's quite sparsely populated. It's very rural up there. And that's where all the wind energy is. It's offshore farms and in some cases onshore, but actually, the vast majority of demand is based down in the South, down in London. So the energy that's being produced up there in Scotland. Needs to find its way down south. 

Now, the problem at the moment is that as far as I understand it, the national grid is not producing data that tells you where that curtailed energy is in more specific, more specificity. There aren't also ways to know which data center then would be in a good position to use that data. I think there's work happening on it. It's developing. But right now, if you were a developer, you couldn't just go along, pull out a load of data and make this happen. So there's work to be done here, in this field. And, you know, as I mentioned already, there's real potential in this field. To work on those things, this is what we wanted to propose as the next version of Carbon Aware, which we labeled as Grid Aware.

Let's get all these amazing minds working on this part of the problem, rather than getting stuck at the first gate, which was, ooh, acknowledging location time shifting is interesting and fun. We need to push this field forward, and we need to do more on this. 

[00:36:30] Gaël Duez: I was laughing a bit when you were explaining this because I had this image of what Sylvie Daumal recently shared on the episode with systemic design. And in systemic designs, I've got this hearsay that says that solutions of today are problems of tomorrow. And that's just so true here. Once again, this is us trying to jump on a very easy and technical solution and very appealing. And I was like 100% in for this solution and then suddenly posing and saying, in which environment are we truly evolving? In which system in which almost a biosphere, an electrical biosphere, I would say electrical sphere. Once again, systemic design is everything. When it comes to sustainability, I'm realizing that there is maybe something that might not have been 100% clear for some listeners if they're not that familiar with cloud operations and energy. Claim for hyperscalers that yes, almost all of them, except maybe one of them, explained that they run 100% on green energy. And I had this wording, so I would say low carbon energy, but that we really need to understand that when a Google, Amazon or Azure or Huawei data center operates, it is actually plugged into the grid, and it's using the energy that is being made available at this very moment. And when they say they produce 100% green energy, it does not perfectly match the energy they consume. So actually they pay. Ismail, am I correct? Do you want to elaborate a bit on what I said?

[00:38:22] Ismaël Velasco: Sure. And I actually want to focus sort of as we close the episode, on four things that every tech practitioner can actually do already to operationalize this well, and that is within our reach. So the first one is to monitor your own net electricity demand. That's the metric you really want to influence. So more than when is this compute job or this computing job running in the grid? The most impactful thing that I can do is use less energy tomorrow than I did today. If I can do that, I will win. If I can't do that because I'm a growing business, can I slow down the rate of my energy demand? So every year, I consume 3% more electricity. Now, I'm still growing, but I managed to grow only one and a half percent every year. Anything that reduces your demand, that's the biggest win of all. That's the thing that we need. So that's a very practical thing. Quantify, monitor, and improve that. 

The second thing is time shifting. And location shifting can work. If you want to find a metric for time-shifting and location-shifting privilege. Low demand, not grid carbon intensity. In other words, run your jobs when the demand is lowest, not when the grid is greenest. They will very often be the same, but not always. You might have the greenest grid at 12:00 in Brazil, but it's not your lowest demand. So adding a bunch of demand might not be good. In Scotland, yes, if the lowest demand is 03:00 a.m. It's probably the greenest. But if you're going to use an API instead of targeting, when is the carbon intensity the lowest target, low demand, then as a cherry on top. If you want, add carbon intensity. But if you target low demand, the chances that you're going to have a positive effect are higher and a negative effect lower. If you just target carbon intensity, it's very random and unlikely. So that's a simple operational right. If you are going to location shift, time-shift, and you're in a good grid like Britain, et cetera, you're not targeting Texas, you're not targeting Azerbaijan. You target good, strong connected grids, then target low demand first. If you do that, that's progress. We talked about lowering demand. I'm using less energy. Great. 

The third one is, how do I add more green energy? One is the curtailed energy, but that one is difficult, and small. In Scotland, the bottleneck is less for computing because you could put your machine there. But the problem in many grids is that, for example, that excess energy cannot travel to the south, where it's being used. So unless you're located there, it's going to break. So how could you add more green power? Which goes to your question. So, one way is generating your own power from your own solar panels, from your own wind turbines. The pattern that we have for electricity historically is hyper concentration on massive power plants. The pattern we've had for computing traditionally is hyper-concentration hyperscalers. But both are distributable. You have distributed electricity and distributed computing. And for the global south. The future is microgrids and distributed computing. And for resilience, they're investing in Texas, because then the whole thing doesn't break if you've got microgrids. So if you were able to power your computer from your microgrid or your solar panel, you are adding extra generation directly. That's brilliant. Has huge potential. And then comes the answer to your question, but what if I run it on Google? Aren't they 100% green? They have done very good things. But what you say is absolutely true. Their data centers are powered by the same grid as everybody else's, which is mixed all the time. So they cannot say my server is actually powered by the sun. They're being powered, but what they do is that they buy what they call additional electricity, which is good. The bad pattern of those things is you're generating green electricity. I'll buy you and I'll take the credit. But actually the demand hasn't changed. Instead, these people are paying someone to build a new wind farm to generate the same energy they will consume. So it's new, it's additional. They can say it is added, but it's not direct. That has also the risk of side effects. So it's not perfect, but it is okay. It's acceptable as long as you are going for additional energy. 

The fourth one is what you are doing. We need to talk about this. For me, the biggest shock in this process has not been the discoveries, but that I haven't found this conversation happening virtually at all. It's been quite shocking to me that there is no very clear point of reference. Even now, since I published the Hackernoon article, I've had talks with some of the very pioneers of these patterns and no one seemed to actually say, is this reducing emissions? Does this have side effects? And I also want to share that someone in one chat wrote, when this came through, I was going to make a post about my concerns with carbon aware software, but I was afraid to do so. There is a lot of commitment to this idea and a lot of investment in this idea, emotional and economic, from both environmentalists and companies. ET was interesting. That's one of the things that most excited me about working with Hannah was that I just thought, I'm nervous. How do I communicate in a way that is clear-cut enough that we can say there is a problem, but that also doesn't sound like I am hating on people? I've been a bit more direct on the hacker noon one because I think we need to be very direct. But basically, above all, I think we need to create a space where we are talking about this and writing about this. I would love to be proven wrong. That'd be even better. But above all, I think nobody should feel a hesitation to say this might be problematic. As I've been following this thread, I am beginning to question some of our carbon emission metrics for software that we've been using. So it's like a little threshold. We're also exploring with the Mayan people here in Mexico, some experiments with distributed energy generation and distributed compute. So we might be piloting those things. There are companies like Soluna in Morocco, there's a company in Paraguay who do exactly this. They go to power plants and they install servers. The servers help the power plants, and smaller plow plants manage their demand response and monetize it. And they're able to sell server space at a fraction of the hyperscalers. So there is some work going on, all of those things, but it's mostly invisible. So if we can agree to demand evidence from industry, just to demand evidence of impact and risk assessment, just those two things. And if we can explore grid aware software or whatever, or carbon aware software 2.0, but just talk about it, we're still on time, I think, for this not to get out of our hands, but it's a small window.

[00:47:10] Gaël Duez: Yeah, I agree. Thanks for mentioning the podcast. But actually, this is really what I'm trying to do here, is having nuanced conversations. And the main thing that we've been trying to achieve with this discussion is not throwing the baby with the bathwater. Usually, I ask my guests to close the podcast with positive news on sustainability, or even better, on green it or digital sustainability. And I think for the very first time, because this is a first, this is the first time I've got former guests coming back in the podcast over. The very first time.

I'm going to be the one answering my own question about how rude it is. But I think what I actually find very refreshing and energizing, and which brings me a lot of optimism, is that I strongly believe that three years ago, we would never have this kind of nuanced in our way to tackling its emissions and having the ability to say, okay, so we are doing things, and we are doing things at a large scale, enough to start to see that the pattern is not white or black and that we can really kick-start some continuous improvement process, which is actually how it works in every industry. You don't do things perfectly right from the start, and you've got this discussion, and even in the orange industry, you can think about all this ongoing discussion about software development practices, the agile methodology, you've got ongoing discussion over and over trying to fine-tune things, et cetera. And for me, starting to see this kind of discussion happening for green software and more generally green, it is very positive because it means that we're getting a scale where we can see the details, the nuances, and that goodwilling people can debate without any prejudice to each other and try to find what will be the best way. So that would be my closing words.

[00:49:18] Ismaël Velasco: I think carbon aware software done well. Time and location shifting are the future. I think that the more we depend on renewable energy, the more integrated will be time and location shifts in computing to what is called demand response. What Hannah was saying about how the grid has to constantly balance. And when you had fossil fuel only, it was very simple because you had complete control. You open the tap, you close the tap. You open the tap, you close the tap. The reason we have curtailed energy and the reason we have all kinds of risks is that we can't open the sun and close the sun, open the wind, and close the wind. So one of the challenges is when you move to 70% renewable energy, 70% of your energy is unpredictable, which means that we, and this is already happening in many places, will have computed and other forms of event-driven electricity consumption integrated into the grid. So every time the sun goes down, there's low demand, we will have more compute, et cetera. So this kind of. I do think that carbon aware time shifting, and location shifting are intrinsic to the future. And the good news is that the only thing where we're doing well in climate action is the energy transition. That's the only bit where we're ahead of the curve, where we're accelerating. So I think that has implications for us. And as long as we ask enough questions to do it responsibly, I think that all the work that's been done on this will prove to have been visionary and very long-lasting as part of the new societies that emerge.

[00:51:20] Gaël Duez: Okay. That was great to have you on the show, Hannah. I'm so happy that you've been back again. Let's make sure that we meet in London. No, that won't be the case, but in Paris for the Green IO Conference. Thanks a lot for joining. It was a pleasure as usual. 

[00:51:35] Hannah Smith: So thanks, thanks so much Gaël for having us on the show. It's amazing to be able to chat this through and it's really nice to speak it out loud because actually so far my involvement in this work has been written. So actually being here and verbalizing it is a really, really fun activity, and it's been really, really useful to do this. And I love the work you do, and I can't wait to meet you in Paris in December. It's a long time to go, but we will, we will catch up with each other then. 

[00:52:08] Gaël Duez: And thanks a lot Ismael also for joining the podcast with your energy, your refreshing perspective, and your humility you didn't come here saying, Hey, this is how it works. And this is the problem, but more, maybe we should ask questions. And this is what I found so far. And I will be so happy to be proven wrong, which is. I sincerely believe in the very basics of a scientific approach. So thanks a lot Ismael for, for, for joining in bringing us all these insights and, have a good day, in Mexico today because it, we call it still quite early for you, and hope to see you soon in the podcast.

[00:52:48] Gaël Duez: Bye bye. 

[00:52:49] Ismaël Velasco: Thank you. 

[00:52:50] Hannah Smith: I'll see you soon. 

[00:52:51] Gaël Duez: Thank you for listening to this Green IO episode. If you enjoyed it, share it and give us five stars on Apple or Spotify. We are an independent media relying solely on you to get more listeners. Plus, it will give our little team, Jill, Meibel, Tani and I, a nice booster.

[00:53:10] Gaël Duez: Today, I had former guests coming back for the first time. And in our next episode, we will have another premiere with a member of the European Parliament among us, Kim Van Sparrentak will join us as well as Max Schulze, the founder of the Sustainable Digital Infrastructure Alliance to talk about regulations in digital sustainability. What are they? What to expect? But also have they been built in the Bristol Maze? And yes, we will also discuss lobbying. And yes, it matters to people outside Europe because of the long arm of European regulations. By the way, Green IO is a podcast and much more. Visit greenio.tech to subscribe to our free monthly newsletter, read the latest articles on our blog, and check the conferences we organize across the globe. The next one is in Singapore on April 18, and you can get a free ticket using the voucher Green IO VIP. Lucky you. Looking forward to meeting you there to help you dear responsible technology purchase build a greener digital world one byte at a time.


❤️ Never miss an episode! Hit the subscribe button on the player above and follow us the way you like.

 📧 Our Green IO monthly newsletter is also a good way to be notified, as well as getting carefully curated news on digital sustainability packed with exclusive Green IO contents.